Re: [PATCH 09/12] use a shell function for AC_CHECK_TYPE

[Ralf Wildenhues]

Hi Eric,

* Eric Blake wrote on Mon, Nov 03, 2008 at 02:16:07PM CET:
> According to Ralf Wildenhues on 11/2/2008 9:04 AM:
> > * Paolo Bonzini wrote on Sun, Nov 02, 2008 at 09:27:42AM CET:
> >> By a completely different approach, would it make sense to switch from
> >> cat to AS_ECHO in order to build the conftest.c files?  That would save
> >> forks for *all* tests...
> > 
> > I don't remember the details, but there were problems with long output
> > and the various echo replacements used in ltmain.sh.  It's grossly ugly,
> > yes, but I did have to debug such a failure once (probably on a Solaris
> > systems with its broken printf), and it wasn't fun.
> 
> Well, gnulib-tool's choice of echo replacements currently causes Solaris
> /bin/sh to dump core, so there are still bugs if you make the wrong
> choice.  But m4sh's choice of echo replacements already tries to detect
> the case of printf that can't handle long string; are we truly robust in
> that area, or just lucky so far that none of our strings are long?

I have no idea.  FWIW, there are still segfaults with ksh that we've
never really gotten to the bottom of, see this thread and follow the
various links in it, for some instances (which may be same or different
bugs):
<http://thread.gmane.org/gmane.comp.sysutils.autoconf.general/8024>

Cheers,
Ralf